Adaptive Machine Learning for Biomarker Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current image analysis technologies in biomedical fields rely heavily on human hand-scoring, which is labor-intensive and prone to variability and error, and lack automated solutions for detecting and quantifying biomarker targets in biological samples, especially for new classes of constituents.

Innovation Solution

A cloud-based, AI-as-a-Service model with computer vision capabilities that uses machine learning and transfer learning for automated detection and quantification of biomarker targets in biological samples, allowing for user interactions to correct annotations and refine algorithms, enabling high-throughput analysis and adaptation to new classes of constituents.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If human hand-scoring is used for image analysis, then flexibility in handling diverse biological samples is maintained, but labor intensity and human error increase significantly

Engineering Contradiction:

ImproveFlexibility in handling diverse samplesVSAvoidLabor intensity and analysis speed

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The system enables automated self-service analysis where the machine learning model independently processes biological images, identifies constituents, and generates quantification results without requiring manual scoring, thereby eliminating labor-intensive human intervention while maintaining operational flexibility

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical human hand-scoring process with an automated computer vision system that uses machine learning algorithms to detect and quantify biological constituents, substituting human manual operations with automated digital processing to improve productivity and reduce errors

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If traditional image analysis methods are used, then existing protocols are followed, but automated detection and quantification of biomarker targets is not achieved

Engineering Contradiction:

ImproveAdherence to established protocolsVSAvoidAutomated biomarker detection capability

Core Design Contradiction:

ReliabilityVSExtent of automation

Solution Approach 1:

The system incorporates feedback mechanisms where user corrections of automated annotations are fed back into the training dataset, allowing the machine learning model to iteratively improve its accuracy and reliability while maintaining automated operation, thus bridging the gap between established protocols and automated detection

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent implements preliminary action by pre-training machine learning models on extensive datasets before deployment, and by providing automated initial annotations that users can review and correct, thereby establishing reliable automated detection capabilities while maintaining protocol adherence through iterative refinement

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If standard machine learning models are deployed, then existing algorithms are used, but adaptation to new classes of constituents is limited

Engineering Contradiction:

ImproveDetection accuracy of known constituentsVSAvoidAbility to detect new classes of constituents

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system implements dynamic adaptability where the machine learning model can be continuously retrained and updated with new data from diverse biological samples, allowing it to dynamically adjust to new classes of constituents while maintaining high detection accuracy for known constituents through iterative learning

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent creates a universal machine learning framework that can detect multiple classes of biological constituents across different sample types and imaging modalities, enabling the system to handle both known and new constituents through a single adaptable platform rather than requiring separate specialized models

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Measurement precision

If manual analysis is performed, then detailed examination is possible, but throughput and scalability are severely constrained

Engineering Contradiction:

ImproveDetail of analysisVSAvoidThroughput and scalability

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system segments the analysis process into distinct automated stages including image preprocessing, constituent detection, quantification, and result generation, allowing each stage to be optimized independently while maintaining detailed analysis capability across high-throughput processing of multiple samples

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11715204B2Adaptive machine learning system for image-based biological sample constituent analysis

Publication Date: 2023.08.01 REWIRE NEURO INC

AI summary

Systems and methods for image-based biological sample constituent analysis are disclosed. For example, image data corresponding to an image having a target constituent and other constituents may be generated and utilized for analysis. The systems and processes described herein may be utilized to differentiate between portions of image data corresponding to the target constituent and other portions that do not correspond to the target constituent. Analysis of the target constituent instances may be performed to provide analytical results.